[Feature] Add temp_scaled_logprobs and top_p_normalized_logprobs parameters for logits and logprobs post processing (#3536)

* [feature] Add temp_scaled_logprobs and top_p_normalized_logprobs parameters for logits and logprobs post processing

* infer engine support temp_scaled_logprobs and top_p_normalized_logprobs

* code check

* code check

* fix tokenizer.decoder(-1), return 'Invalid Token'

* check seq len time shape

* logprob clip inf

* code check

---------

Co-authored-by: sunlei1024 <sunlei5788@gmail.com>

fastdeploy/entrypoints/openai/protocol.py

@@ -333,6 +333,9 @@ class CompletionRequest(BaseModel):
     echo: Optional[bool] = False
     frequency_penalty: Optional[float] = None
     logprobs: Optional[int] = None
+    # For logits and logprobs post processing
+    temp_scaled_logprobs: bool = False
+    top_p_normalized_logprobs: bool = False
     max_tokens: Optional[int] = None
     n: int = 1
     presence_penalty: Optional[float] = None
@@ -461,6 +464,11 @@ class ChatCompletionRequest(BaseModel):
     frequency_penalty: Optional[float] = None
     logprobs: Optional[bool] = False
     top_logprobs: Optional[int] = 0
+
+    # For logits and logprobs post processing
+    temp_scaled_logprobs: bool = False
+    top_p_normalized_logprobs: bool = False
+
     # remove max_tokens when field is removed from OpenAI API
     max_tokens: Optional[int] = Field(
         default=None,
@@ -515,6 +523,8 @@ class ChatCompletionRequest(BaseModel):
 
         req_dict["max_tokens"] = self.max_completion_tokens or self.max_tokens
         req_dict["logprobs"] = self.top_logprobs if self.logprobs else None
+        req_dict["temp_scaled_logprobs"] = self.temp_scaled_logprobs
+        req_dict["top_p_normalized_logprobs"] = self.top_p_normalized_logprobs
 
         # parse request model into dict, priority: request params > metadata params
         if self.metadata is not None: